Puget Systems print logo
Read this article at https://www.pugetsystems.com/guides/1123
Article Thumbnail

After Effects CC 2018 CPU Comparison: Intel 8th Gen vs X-series vs Ryzen vs Threadripper

Written on March 8, 2018 by Matt Bach


Since AE 2015, choosing the right CPU has been a bit of a challenge as exactly the right CPU has been subtly and steadily changing. In AE 2014 and before, the "Render Multiple Frames Simultaneously" feature made it simple: more cores is better. However, with AE 2015 this feature was dropped as it was incompatible with the new GPU acceleration that Adobe was developing for After Effects. This change largely made core count secondary to raw CPU frequency, but things have again been changing over the last year or so due to the integration with Cinema4D.

To be fair, for most users a high frequency CPU like the Core i7 8700K should still give them the best performance in After Effects - easily outperforming more expensive CPUs that are up to 4x the cost. However, if your projects include 3D Rendering, Adobe has been steadily shifting over to a workflow that takes advantage of the recent integration with Cinema4D. The older "Classic 3D" and "Ray-traced 3D" renderers are still available, but from what we can see Adobe largely halted development of these several years ago. While not based on any information from Adobe, in our opinion there is a decent chance that these older renderers will be dropped completely in future versions, making C4D the only way to get 3D Rendering in After Effects.

What it comes down to is that there are two workflows in AE that each prioritizes a different aspect of the CPU: "standard" projects which favor CPU frequency and projects using Cinema4D which should favor higher core counts. To see how various CPUs compare in each of these workflows, we will be looking at RAM Preview and Final Render performance with a range of After Effects projects. If you would like to skip over our test setup and analysis of the individual benchmarks, feel free to jump right to the conclusion section.

Test Setup

For After Effects, there is a wide range of CPU options that might be great depending on exactly what you do. For general AE tasks, a CPU with high single-threaded performance and a moderate core count like the Core i7 8700K should give you the best possible performance. Depending on how much you use the newly integrated "Cinema 4D" 3D renderer, however, you might see massive performance increases with a higher core count CPU like the Intel X-series or AMD Threadripper CPUs.

Because of the wide range of useful CPUs, we decided to include a range of models in our testing covering various types of CPUs that we felt might be a good choice for After Effects. This includes choices from the Intel 8th Gen Core i5/i7, Intel X-series Core i7/i9, AMD Ryzen, and AMD Threadripper lines of processors.

On thing we want to note is that we are technically overclocking the Ryzen platform by using DDR4-2666 memory since we are using four sticks of dual rank RAM. According to this blog post - which we have confirmation from AMD that it is still accurate even with the newer AGESA BIOS - the highest RAM speed that is officially supported with four sticks of dual rank RAM is just DDR4-1866. Our initial Ryzen testing was performed with DDR4-2400 RAM, but even then we received a lot of comments about how we were artificially limiting the performance of Ryzen even though we were actually overclocking the platform. Using DDR4-2666 RAM doesn't really change the results too much, however, so we opted to go ahead and use the faster RAM. We're sure some will say we should use even faster DDR4-3200 RAM, but there is a limit to what we are willing to use considering our testing is first and foremost to help ensure we are offering the right hardware to our customers. Since DDR4-3200 is well beyond what we consider to be acceptably stable for most end-users, it is extremely unlikely to be a part of our testing in the near future.

We also want to point out that while our test platforms are using a single hard drive, that is not actually what we would typically recommend to our customers. A two (or more) drive configuration - with the cache files on a secondary drive - can make a big impact when scrubbing through complex projects. But since we will not be testing anything that would benefit from multiple drives in this article, we opted to use a single drive to cut down on the number of hardware variables.

In order to accurately benchmark the different CPUs, we used a range of After Effects projects that are mostly publicly available for download. We did this so that anyone could replicate our testing on their own computer to see just how much of a performance boost they could expect if they upgraded to one of these CPUs.

The projects we used (along with their source) are:

Project Name Comp Name Tested Frames Notes
by Rocketstock

Example Composition 0-40 (40 total frames)  
Electric FX
by Video Copilot

212-238 (26 total frames)
0-97 (97 total frames)
by EditingCorp

Main Comp 0-200 (200 total frames)  
Animated Polygon
by Video Copilot

Green Polygon 0-78 (78 total frames)  
GPU Stress
  32610-32710 (100 total frames)

4K H.264 video with:

  • Find Edges
  • Glow
  • Brightness/Contrast
  • Transform
  • Sharpen
  • Directional Blur
  • Lumetri Color
Cineware Party
by Cineversity

Party-Lite-004Full 0-169 (169 total frames)

"Video Wall" and "*.mov" layers
removed. Tested with C4D Renderer:

  • OpenGL
  • Standard (Draft)
  • Standard (Final)

RAM Preview (RAW DATA)

[Click Here] to skip ahead to analysis section

RAM Preview (Analysis)

More than almost anything else, high performance during RAM Preview is one of the most critical aspects of a productive After Effects workstation. However, with the integration of the Cinema4D rendering engine (which should be able to utilize a higher number of CPU cores), the best CPU choice will often depend on whether or not you make use of this new feature. Because of this, we will divide up our analysis between "Standard" projects and those that utilize the CPU render portion of the Cinema4D rendering engine.

After Effects 2018 RAM Preview CPU Benchmark
Starting with the "Standard" projects, we see performance that is almost identical to many of our previous articles. In fact, compared to our After Effects CC 2017.2 CPU Performance: Core i7 8700K, i5 8600K, i3 8350K article, the relative performance between these CPUs is effectively the same. There are minor differences due to the fact that we have recently updated our test projects, but overall it is clear that the Core i7 8700K is still king due to it's high single-threaded performance - although the Core i7 8700 is only a tiny bit behind.

After Effects 2018 Cinema4D Rendering CPU Benchmark
One of the changes we made with our new set of test projects was to include one that used the Cinema4D rendering engine. In our previous article we "faked" this kind of testing by using the CineBench benchmarking utility, but if you compare those results to what we saw in this real-world test you will notice that the results are dramatically different. Higher core counts are in general still better, but the AMD CPUs in particular did not do very well. For example, while CineBench reports the AMD Threadripper 1950X should be roughly twice as fast as the Core i7 8700K, when you are using C4D through After Effects that CPU is actually anywhere from 23% slower to just 16% faster than the 8700K depending on the playback resolution. This is an absolutely massive difference and really goes to show how important real-world testing is rather than relying on stand-alone benchmarking utilities.

This lower performance is also apparent with the Intel CPUs, although it is not nearly as dramatic. In fact, we actually hit a wall at the Core i9 7940X 14 Core where using even higher core count (and more expensive) CPUs resulted in a loss of performance. This is likely due to the fact that those CPUs run at a lower frequency and After Effects really relies on a high CPU frequency for optimum performance.

Final Render (RAW DATA)

[Click Here] to skip ahead to analysis section

Final Render (Analysis)

After Effects 2018 RAM Preview CPU Benchmark
For Final Render performance, we once again decided to break out the projects using the CPU rendering portion of Cinema4D since whether or not you use that feature dramatically changes performance.

Starting with the "Standard" projects, the relative performance between each CPU is pretty much identical to what we saw in the RAM Preview testing when we used full resolution preview. This makes sense as rendering each frame into RAM is really not all that much different than rendering each from to the hard drive even if there is the addition step of encoding it to a .AVI video file.

In fact, this holds true for the projects that use Cinema4D CPU rendering as well. The AMD CPUs did a bit worse than our RAM Preview test at full resolution, but otherwise pretty much everything we talked about in the previous section completely holds true for this test as well.


After Effects 2018 CPU Comparison Benchmark

Averaging the results from both the RAM Preview and the Final Render tests - but keeping the Cinema4D CPU projects separate - we get a clear look at how the various CPUs we tested compare in After Effects 2018.

The first thing we want to call out is how different the Cinema4D CPU rendering results are compared to what we saw in our previous article where we simply used the CineBench benchmark utility (chart from that article is to the right). CineBench is an excellent benchmark, but we really should have anticipated how much After Effects itself would affect performance. Relying on that benchmark utility alone is led us to recommending AMD Threadripper for those that are taking advantage of the new C4D integration, but that platform is clearly not a great choice. So before anything else was want to apologize and retract our previous recommendation. Based on our new testing, If you are using C4D through After Effects we would recommend the Core i7 8700K 6 Core as a great starting point followed by either the Core i9 7900X or Core i9 7940X if you need higher performance.

For general AE usage, there are pretty much two types of CPUs that you might choose. To get the best performance overall performance, it is going to come down to one of the Intel 8th Gen CPUs like the 8600K, 8700, or 8700K. The Core i7 8700K is currently our go-to recommendation, although the Core i7 8700 is only very slightly behind for a small savings in cost. The downside to these CPUs is that they are limited to 64GB of RAM. This should be plenty for most users, but if you tend to store a lot of high resolution frames in RAM Preview than one of the X-series Intel CPUs may be a better choice.

If you are one of those users that need more RAM, these X-series CPUs support up to 128GB of RAM but they are ~10% slower for most AE projects compared to the Core i7 8700K. However, having that RAM is going to more than make up for that small performance loss for those that can really utilize it as it will all you to play through more frames without the potential need to re-render them over and over.

Whether a Core i7 8700K or one of the higher-end X-series CPUs is right for you is going to entirely depend on both your budget, where you need more performance, and how much system RAM you need. Overall, however, the Core i7 8700K in particular is an excellent overall CPU for After Effects while the Core i9 7900X and Core i9 7940X are great for those that need more system RAM or those that are making heavy use of the new integration with Cinema4D.

After Effects Workstations

Puget Systems offers a range of poweful and reliable systems that are tailor-made for your unique workflow.

Configure a System!

Labs Consultation Service

Our Labs team is available to provide in-depth hardware recommendations based on your workflow.

Find Out More!
Tags: After Effects, Coffee Lake, 8th Gen, X-Series, Ryzen, Threadripper, Cinema4D
Gabriel Puglisi

Hi, sorry for my english, but can you tell me more about the performances needed to work on task like Character animation with Duik, Puppet, Rubberhose etc...?
With this tools, that I use a lot, i'm very often confronted to a very slow down workflow, in RAM preview and renders. And of course, more I have characters, more it's slow. I think it's due to the high number of expression and the fact there is a lot of expression to interpret on each frame.

I have a 64gb RAM and 7700k I7 configuration, SSD 960 EVO for the cache and a GTX980ti.
I would like to buy a better / faster system, but if I understand well yours benchmark, a CPU with more cores will not help for common works.
But you do not tell about Character animation...

So I'm little bit confused, do you think it will help for this task to buy a 128Go RAM and 7900K configuration for example ?

Thanks for your help.

Posted on 2018-03-09 17:12:54

compra el 8700k, 3 discos M2 960 Pro de samsung, 1 sietema, otro proyecto y otro cache de after, y 64gb de ram lo mas rapida que alcanze a comprar tu presupuesto junto con la placa base compatible a el tipo de ram que comp[raras, es lo mejor para lo que quieres, tengo un threadripper 1950 x y lo que te comento y es frustrante trabajar con la suite de adobe en especial after effects, ve a lo seguro y no tires tu dinero, espero que te ayude saludos!

Posted on 2018-04-02 22:19:12

How much cores in Ryzen and 8700k was loaded during rendering&creating previews? I am wondering because if you have TR 1950X you can keep some resources for another programs. Can you receive the same performance with 12 or 10 cores enabled in TR?

I did small research with Premiere and my TR. There are no difference in render speed between 16 and 10 cores... and 12 cores was actually faster (lol). and 8 cores was slower than 16 just about 5%. The weirdest thing is all cores was loaded near 100% every time

Posted on 2018-03-13 20:53:04

I have a proposal that will make things more productive in articles of this type.

1) All performance scores that are presented as a production rate (frames or MB per second) should stay as they are today. "Higher is better."
2) All performance scores where the result is a time ("lower is better"), should be presented as a production rate, as solutions per hour. Higher will be better. So, it's not 30 seconds, 120 solutions/hour. Higher is now better in all graphs.

Seriously. It is this simple. It should be a standard. Countless hours will be saved in preparing and reading graphs. No more higher is better and lower is better text. No more mistakes in reading graphs. No more confusion.

In youtube videos where graphs stay on the screen for only a few seconds, the situation with the lower is better graphs is getting ridiculous. Please apply common sense, people!

Posted on 2018-03-14 11:45:19

I personally disagree, as I find that the context of a test often makes 'lower is better' the best way to look at results, but I respect your opinion and appreciate your input :)

Posted on 2018-03-14 15:55:08

In engineering, it's usually time/unit for lengthy tasks, and units/time for every other productivity measure. I would like to know that a task takes 2 hours instead of 3 hours, because that means a lot in a productive environment, personnel time, overtime, working hours, scheduling and delivery time, etc. When I read a test like this though, where the tasks are only a few frames, I usually want to quickly know that it's a lot faster in task A, and a llttle faster in task B or the actual productivity in fps.

A typcal example for this article: Final Render 0 Polygon (78 frames), render time (seconds) - Lower is Better.
AMD TR 1950x 31.8 sec
Intel i5 7600K 27.4 sec

I think the following is more useful for all purposes:

AMD 2.45 fps
Intel 2.85 fps

All the time graphs in this article are average fps really, not different from fps in a game.

Units/time (production rate) is also easier to interpret. 2.45fps means that it's basically 10x real time for my 24p task. 1 minute will take 10 minutes. 31.8 seconds for 78 frames is not as easy to interprate.

Also, in other tasks on this article, we have a lot more frames in the test. If the results were presented in fps, we would be able to directly compare productivity in different tasks. That may or may not be useful, but it certainly doesn't hurt.

Posted on 2018-03-15 04:25:46

An example from the industry: A lot of time was put on designing SpecViewperf. It's not reporting time for N frames or frames in time X (which are the things benchmarks actually measure), it converts to fps and reports fps, the production rate of the system. We can all relate to production rate. It's a very good measure of responsiveness and performance.

Posted on 2018-03-15 04:55:47

An example from the car industry. Consumption is actually gallons/mile, but most people want to see a production rate because that's easier to interpret, so consumption is presented in miles/gallon. High consumption is worse, but this way higher is better and the more productive car will score higher. They certainly would not want to see consumption in gallons for an arbitrary distance, or the distance covered with an arbitrary quantity of fuel.

Car emissions on the other hand, are not considered productive in any way, so higher is worse, so the manufacturers state them as they are calculated. These are not really a measure of performance though, they are only used in judging if they are within limits for tax reductions such as the ones applied in Europe in cars with low emissions.

But if the customers are only doing long tasks once or twice, such as baking pizza, the pizza oven manufacturer will tell them it can produce a pizza in 14 minutes. The professional will still get a pizzas/hour performance metric (tunnel type pizza ovens are very fast). Different scenario of use.

There is usability aspect in everything.

Posted on 2018-03-15 05:16:00

Things look different for high core count CPUs if you start using RenderGarden Plugin for AE...

Posted on 2018-03-23 11:04:34

Seriously i spent $210 on a new 8700, i can buy other 4 and i will have 4 times better performance than a $2000 cpu for $800... :-)

Posted on 2018-03-31 01:11:19

almost, i was thinking same as u... but... then i realize that i will need too another pc components kit, i mean its not just 1 cpu, is 1 mobo, rams, cooler, ups, case, video card, etc etc etc, anyways, u can have 2 pcs, one for work and another for render while u work

Posted on 2018-04-02 22:25:28
James Nielson

Pardon my ignorance, @PugetSystems, but how do you test RAM Preview time? A stopwatch, or is there something fancier I don't know about?

Posted on 2018-07-24 02:43:32
James Nielson

Also, I think I speak for quite a few people when I say we'd love to see the 8750h/8850h/8950hk benchmarked and compared.

Posted on 2018-07-24 02:45:44

Basically just a glorified stopwatch really. We are using After Effects scripts to do the actual timing but it is basically: start timer & playback (single pass, no loops), stop timer when playback is done, log both the time and the number of frames played (more for making sure nothing changed that shouldn't have), then calculate the FPS based on the time and number of frames.

As for testing mobile CPUs, that is not something we are likely to test to be honest. We used to sell laptops several years ago, but dropped them in order to focus on workstations and servers. In the end, what we test is primarily based on what we need to know to ensure that we are selling the exact right things to our customers and while we do sometimes go outside of that (like with our Mac vs PC testing), testing mobile CPUs isn't very high on the list. I appreciate the feedback though - if we get enough requests for something outside our normal content we can sometimes make it happen.

Posted on 2018-07-24 16:25:04
Mohammad Basel

Folks... what would be the most powerful/ best laptop choice to run after effects at its best resolution?
I am considering the ASUS Gx800vh, its expensive but i need the best possible laptop in the world to buy because
my clients are based in different cities, i would really appreciate it if you can point me in the right direction to purchase the laptop right away.

Posted on 2018-08-16 14:09:25
Prabhu Ram

This is what I was exactly looking for.

Posted on 2018-09-08 12:44:19